Você está na página 1de 8

Computers & Geosciences 63 (2014) 6269

Contents lists available at ScienceDirect

Computers & Geosciences

journal homepage: www.elsevier.com/locate/cageo

A numerical model of ballistic transport with collisions

in a volcanic setting
Kae Tsunematsu a,c,n, Bastien Chopard b, Jean-Luc Falcone b, Costanza Bonadonna c

Graduate School of Environmental Studies, Nagoya University, Nagoya, Japan

Department of Computer Science, University of Geneva, Geneva, Switzerland
Section of Earth and Environmental Sciences, University of Geneva, Geneva, Switzerland

art ic l e i nf o

a b s t r a c t

Article history:
Received 1 July 2013
Received in revised form
19 October 2013
Accepted 23 October 2013
Available online 30 October 2013

Fragments associated with explosive volcanic eruptions range from microns to meters in diameter, with
the largest ones following ballistic trajectories from the eruptive vent. Recent eld observations suggest
that bombs ejected during Strombolian eruptions may collide while airborne. We developed a Discrete
Event Simulator to study numerically the impact of such collisions on hazard assessment. We show that
the area where bombs can land might be signicantly increased when collisions occur. As a consequence,
if collisions are dominant, the deposition distance cannot be used to estimate important eruption
parameters, such as exit speed.
Crown Copyright & 2013 Published by Elsevier Ltd. All rights reserved.

Volcanic eruption
Ballistic transport
Discrete Event Simulation (DES)

1. Introduction
When a volcano erupts explosively, volcanic particles of various
sizes and shapes are ejected from the vent. Large particles
( 4 64 mm) are transported in the air, decoupled from the gas
phase at the early stage of transport, and follow independent
parabolic trajectories. These large particles are called ballistic
blocks or bombs, or simply ballistics. Their study is crucial as the
kinetic energy and high temperature associated with ballistics can
signicantly damage infrastructures in proximal areas. In addition,
the behavior of bombs allows a better understanding of the
dynamics of ejection because their velocity is more strongly
related to the jet phase than to the convective phase of a
volcanic plume.
There are many numerical models describing the trajectory of
the individual particles. A numerical model of volcanic bombs was
rst suggested by Wilson (1972) which accounted for the effect of
drag forces. It was further developed by Fagents and Wilson (1993)
with realistic conditions including the coupling between the
explosion and surrounded air for Vulcanian eruptions. Bower
and Woods (1996) considered ballistic trajectories which were
accelerated by the gas phase, while taking into account the
particle motion through the crater and the atmosphere.

Corresponding author at: Graduate School of Environmental Studies, Nagoya
University, Nagoya, Japan.
E-mail address: kae.tsunematsu@gmail.com (K. Tsunematsu).

Alatorre-Ibargengoitia and Delgado-Granados (2006) measured drag force of ballistic particles due to the shape difference
in laboratory experiments. The effect of fragmentation on the
initial velocity of ballistics in Vulcanian eruptions has been
included. The results were applied to the hazard map of Popocatpetel volcano (Alatorre-Ibargengoitia et al., 2012).
Mastin (2002) proposed a Graphical User Interface (GUI) with
which researchers can investigate their own scenario. Saunderson
(2008) suggested equations of motion with the resistance and total
centrifugal terms by using plural particles. Recently, de'Michieli
Vitturi et al. (2010) suggested a Lagrangian model of large volcanic
particles subject to a drag force and the background ow eld of
the plume. However none of these models consider the effect of
inter-particle collisions.
Recently Vanderkluysen et al. (2012) reported collisions
between particles. Particle trajectories were extracted from thermal video images of Strombolian eruptions. According to these
trajectories, 12% of the analyzed particles experienced a collision
with another particle. This suggests that inter-particle collisions
should be implemented in numerical models aimed at predicting
the trajectories of bombs.
In this paper, we propose a numerical model that includes
inter-particle collisions. We investigate their effect on pyroclast
travel distances, which is directly related to hazard assessment.
In order to highlight the role of collisions in the ballistic process,
we neglect the drag force. This allows us to use a Discrete Event
Simulation (DES) method, which is a fast numerical approach.
The parameters of our model are based on observations of
Strombolian eruptions. In addition to the thermal video images

0098-3004/$ - see front matter Crown Copyright & 2013 Published by Elsevier Ltd. All rights reserved.


K. Tsunematsu et al. / Computers & Geosciences 63 (2014) 6269

from Vanderkluysen et al. (2012), the initial velocity of particles,

their ejection angles and the frequency of bursts can be extracted
from the work of Chouet et al. (1974), Blackburn et al. (1976),
Ripepe et al. (1993) and Patrick et al. (2007) where photos of short
exposure time are analyzed to obtain the trajectory of particles.
Ejection velocity, maximum height of particle during the ight and
eruption durations can be obtained from Patrick et al. (2007)
where thermal video (Forward Looking Infrared Radiometer)
studies are reported.
In the present work, we consider a sensitivity analysis of the
ballistic process with respect to collisions. We study which
parameters affect most the probability of collision, and which
parameters inuence the travel distance of the bombs when
collisions are taken into account. Our main result is that the area
in which bombs are likely to fall may signicantly increase when
collisions occur.

2. The model

The ejection speed is dened by its magnitude v and its direction

is given by three angles, (rotation), (inclination) and
(azimuth), as shown in Fig. 1. The rotation angle accounts for
the fact that the vent orientation, or the conduit shape, may depart
from the vertical direction. Assuming that the x-axis is chosen so
that the xz-plane contains the vector normal to the vent, the
velocity vector v e is obtained as
0 sin
v sin cos
! B
0 C
ve @
A@ v sin sin A


v cos

The above parameters are summarized in Table 1. In our

simulation, their value, for each event and each particle, is chosen
from a probability distribution to reect their uncertainty. The
types of distributions are shown in Table 2 and have been
determined from the observations reported in Vanderkluysen
et al. (2012), Chouet et al. (1974), Blackburn et al. (1976) and
Patrick et al. (2007).

2.1. Initial parameters

2.2. Ballistic movement and collisions

Our numerical model simulates an eruption as a series of

successive bursts. A burst is the simultaneous ejection of Np
spherical particles of radius R and density p. Bursts are events
that are instantaneous and repeat every time interval t b , as
illustrated in Fig. 1. During a burst, Np particles are ejected at once.
The time of the rst burst is dened as t 0. The time of the last
burst is denoted Tl, an input parameter of the model. The number
of bursts Nb is then given by T l =t b . The total number of particles
N that are ejected during the full simulated eruption is therefore
N Nb  Np .
Each ejected particle is characterized by its offset position r e
with respect to the center of the vent, and its initial velocity v e .

Once a particle has been ejected from a given vent location,

with a given initial velocity, it follows a parabolic trajectory r t,
due to gravity, until it reaches the ground or collides with another
airborne particle. In the present study we disregard the effect of
the drag force. One reason is to better highlight, in a simple
situation, the role of collisions compared to the situation without
collisions. Secondly, neglecting the drag force allows us to signicantly speed up the simulations by using the DES approach
discussed in Section 2.3. Computation speed is important as we
need to perform a large number of simulations to sample the
space of possibilities.
The landing position of a particle on the ground is obtained by
solving the equation G r t 0, where Gx; y; z 0 is the relation
describing the ground prole. If Gx; y; z is a simple function (e.g.
Gx; y; z z z0 as we use here), the time of the impact and its
location can be computed analytically.
To solve the collision process, we use standard classical
mechanics. Let us denote by r i t the trajectory of particle i, with
radius Ri. The pair of particles (i, j) that will collide rst is
determined. This is the pair for which the solution of
j r i t r j tj rRi Rj yields the smallest time t which is not
smaller than current time.
This time t can be computed analytically for any pair of
parabolic trajectories. For instance, let us consider particles
1 and 2:
!0 !0
r 1 t r 1 v 1 t

t 1 12 g t

t 1 2

!0 !0
r 2 t r 2 v 2 t

t 2 12 g t

t 2 2

where ti is the time at which particle i is ejected from position r i
with velocity v i , and g 0; 0; 9:81 is the gravity. It is then
easy to show that r 1 t r 2 t is a linear function in time:
r 1 t

! !
r 2 t a b t

Fig. 1. Input parameters. (a) Rotation angle (), inclination angle () and azimuth
angle (). (b) Displacement of ejection point from the vent center r e x0 ; y0 ; 0.
(c) Parameters related to bursts such as number of bursts (Nb) and number of
particles per burst (Np). Burst time interval t b is the time between tb0 and tb1
(t b t b1 t b0 ). Time of last burst Tl is given deterministically and the number of
bursts Nb is controlled by burst time interval t b .

! !0
a r1


v 1 t 1 v 2 t 2 12 g t 21


g t 1

t 22

! !0
b v1



K. Tsunematsu et al. / Computers & Geosciences 63 (2014) 6269

Table 1
Parameter notation, distributions and inputs. represents average and s represents standard deviation. N ; s indicates normal ( Gaussian) distribution dened by the
mean and the standard deviation s. Note that when a value out of range (e.g. p o 0) is generated, it is re-drawn. U(min, max) indicates uniform distribution between the
minimum min and the maximum max.



p , sp

Particle density (kg/m )

p N p ; s2p

Particle diameter (m)

D 2R
v j v ej

D N D ; s2D

N 0; s
U0; 2
x0 N 0; s2r

Magnitude of initial velocity (m/s)

Rotation angle (deg., rotation from vertical axis)
Inclination angle (deg.), from vertical axis)
Azimuth angle (deg.)
Displacement of ejection points from the vent center (m)

D , sD
v , s v

v N v ; s2v

Deterministic A 0;


Number of particles per burst

r e @ y0 A

N p N Np ; s2Np

N p , s N p

Number of bursts
Total number of particles
Burst time interval (s)

t b

N b T l =t b
N Np  Nb
t b N t b ; s2t b

t b , stb

Time of last burst (s)


Deterministic T l A 0; 1


Table 2
Values of input parameters for our analysis (from Vanderkluysen et al., 2012;
Chouet et al., 1974; Blackburn et al., 1976; Patrick et al., 2007).





Particle density


Particle diameter

Magnitude of initial velocity



Standard deviation of inclination angle

Rotation angle
Displacement of ejection points
from the vent center
Number of particles per burst




N p
sN p

Burst time interval

t b

Time of last burst

st b

As a result, the condition

r 1 t

r 2 t



R1 R2 2 becomes

the quadratic equation:


! ! !2
R1 R2 2 2 a  b t b t 2 0

which is easily solved analytically.

Note that since we do not discretize the time variable t, the
probability to have a triple collision, or several simultaneous pair
collisions, can be neglected.
The collision between the two selected particles is supposed to
be instantaneous. Conservation of momentum, combined with the
fraction of energy lost in the collision (given by the restitution
coefcient CR), is needed to compute the new velocity of the two
particles after collision.
Let us call v 1 and v 2 the pre-collision velocities of the two
particles. It is then well known (Landau and Lifshitz, 1976) that the
problem can be solved by projecting the velocities on the unit
vector e connecting the center of the two particles and the
perpendicular direction. It can be shown that the following
relations denes the post-collision velocities v 1 and v 2 :
! !
v 1 V 1 e v 1

! !!
v1 ee

! !
v 2 V 2 e v 2

! !!
v2 ee

y0 N 0; s2r

! ! ! !
v1 e v2 e ! !
v1 e
m1 =m2 1
! ! ! !
1 C R v 2  e v 1  e ! !
v2 e
m2 =m1 1
V 1

1 C R

2.3. Discrete Event Simulation method

The above model can be efciently implemented using a DES
method (Banks and Nelson, 2010). The DES approach is very
commonly used in the modeling and simulation of systems
characterized by events that occur at well dened moment in
time. Once such an event has occurred, the time of the next one
can be computed precisely from the knowledge of the current
Technically, a DES simulator consists of a state and a queue
of future events. A state consists of the current simulation time t
and the values of the variables describing the system at this time.
Events in the queue are processed until the queue becomes
empty. When an event is processed, all states are updated and the
events in the queue are modied by removing existing events or
creating and inserting new events. Simulation time jumps from
the time of one event to the time of next event. The list of events in
the queue is always kept sorted by increasing time of occurrence,
so that the next event to be processed is the rst in the queue.
In our ballistic model, the state of the system is composed of
(1) the current time, (2) the list of airborne particles, and (3) the
list of deposited particles. Within these lists, each particle includes
its density, diameter, position and velocity.
There are three types of events in our model: Burst (ejection of
Np particles), Collision and Deposition. Between events, trajectories are calculated analytically and new collision and deposition
events can be obtained.
At the initialization, the event queue is lled with exogenous
events, i.e. events which are caused by the external effects (here
the burst). On the contrary, events which are caused by the
internal evolution of the system are called endogenous events
(here the collision and deposition).
At the beginning of the simulation all the particle ejections
(burst events) are inserted in the event queue. Then all the
collision and deposition times are computed according to the
analytical formulas. All these events are added to the event queue
while preserving the time ordering.

K. Tsunematsu et al. / Computers & Geosciences 63 (2014) 6269


When a deposition takes place, the particle is moved from the

airborne list to the deposition list, with the proper time and
location of the event. The event is discarded from the list. The
simulation jumps to the next event.
When collisions take place, the position and velocities of the
two colliding particles are modied according to the momentum
conservation law and the value of the restitution coefcient. Based
on these new velocities, the collision and deposition events
involving the two particles are updated. This means discarding
the collisions which were planned if these two particles had not
collided, and adding new collisions that may occur between these
two particles and all the others.

3. Results and discussion

3.1. Impact of collision on the deposition area
Our model can simulate ballistic transport with or without
collisions. Fig. 2(a) shows an example where collisions are ignored,
even if particles get close enough. Fig. 2(b) shows a similar
situation where collisions are enabled. We can observe that, with
collisions, the traveled distance of the particles is increased.
This is conrmed by Fig. 3 which shows the spatial distribution
of the deposited particles on a at ground prole. The black circles
divide the area around the crater into two regions where 75% and
99% of the particles have landed when collisions are enabled (red
triangle). Green squares correspond to the ground deposition of
the particles when collisions are disabled. The parameters of this
simulation are those in Table 2, except for the particle diameter
which is xed to 1 m and the total number of particles is N 2000.
A typical scenario that explains the increase of the traveled
distance is shown in Fig. 4. A small particle from a previous burst is
falling back to the ground when it is hit by a larger particle just
ejected during a more recent burst. As a result, the small particle is
strongly deviated from its initial trajectory. Similar situations may
also occur between particles of the same mass, as is the case in
Fig. 3, but yet with a lesser effect than shown in Fig. 4.
3.2. Parameter study
The distribution of impacts on the ground when collisions are
allowed depends on the parameters of the simulation. In this
section we investigate the reasons that make particles y farther
away when collisions occur. We consider simulations with the
parameters given in Table 2, except for the parameter whose
importance is investigated.
Total number of particles: The role of the total number N of
particles is considered in Fig. 5. To modify N, the number of bursts
is increased but Np, the number of particles per burst, is still given
by the normal distribution of mean Np and standard deviation
sNp . Fig. 5 shows, for each value of N, the distance to the crater
where a given percentage of the particle has been deposited. The
cases with and without collisions are indicated. In the latter case,
the spatial distribution of the particle is independent of N as
As already noticed, when collisions are allowed, the spatial
distribution of the ballistic impacts is expanded. A reason is that
the total number of collision increases with N and the traveled
distance tends to increase with the total number of collisions.
However, this is not so simple and we will see below that a particle
experiencing too many collisions does not travel very far. What is
important to enlarge the deposition area is to have many particles
that experience only one or two collisions while airborne.
Burst time interval: The deposition distance of the collisionallowed case is sensitive to the time interval between bursts. This

Fig. 2. Illustrations of the ballistic trajectories of 1000 particles for (a) an eruption
without collisions and (b) an eruption with collisions. Different particles are shown
with different colors. (For interpretation of the references to color in this gure
caption, the reader is referred to the web version of this paper.)

dependency is absent when collision is disregarded. Fig. 6

(a) indicates the spatial distribution of the deposits. For short time
interval t b , the total number of collisions is important, as shown in
Fig. 6(b), thus explaining the increase of the deposition range.
If t b is increased to about 10 s, the deposition distances
observed with and without collisions become almost the same.
The number of inter-burst collisions also goes to zero for such
values of t b . This suggests that the important events to increase
the travel distance are the collisions between particles ejected
with a time lag, as in the scenario of Fig. 4.
It is easy to explain why, for time intervals larger than 10 s, the
inter-burst collisions disappear. In our simulation we choose the
average ejection speed to be ve 40 m=s. Thus, the expected
deposition time td of a particle with such an initial velocity is
t d 2ve =g 8:15 s, where g is the gravity acceleration. Therefore, if
bursts are separated by about 10 s, almost all particles will have
landed before the next burst.
In Ripepe et al. (1993) the change of mass ux as a function of
time is reported for a Strombolian eruption. If the frequency of the
change of mass ux is interpreted as the frequency of the bursts,
the burst time interval of ballistics in the 19851986 Stromboli
eruption is around 0.52.0 s. This is clearly shorter than the travel
time of a particle ejected with an initial velocity of 40 m/s.
Therefore, it is possible to have collisions between particles in
successive bursts of a Strombolian eruption.
Additionally, Vanderkluysen et al. (2012) reported that 12% of
particles experienced inter-particle collisions in Stromboli 2008
eruption. It means that bursts occurred from 0.5 to 1.0 s of time
interval according to the results shown in Fig. 6(b). This number
agrees with Ripepe et al. (1993) as discussed above.


K. Tsunematsu et al. / Computers & Geosciences 63 (2014) 6269

Particle size: The effect of particle size on the total number of

collisions and travel distance is demonstrated in Fig. 7. The
number of collisions increases monotonically with the particle
diameter (see Fig. 7(a)). As the particle size increases, the travel
distance increases until a diameter of 150 cm. Then, the travel
distance decreases (see Fig. 7(b)). We suggest that as the particle
diameter increases, the number of collisions each particle experiences increases. As shown below, this is a factor limiting the
traveled distance.
Restitution coefcient: The restitution coefcient CR is a parameter that changes the velocity after the collision due to the

Fig. 3. Particle distribution on the ground (particle diameter 1 m, 2000 particles).

Symbols correspond to the deposition of single particles. Green squares indicate
particle deposition when collisions are ignored. Blue circles indicate deposition for
particles which did not collide even when collisions are allowed. Red triangles
indicate deposition of particles having experienced a collision. Contours indicate
the percentage of deposited particles within the contour: solid lines correspond to
the case where collisions are allowed, and dashed lines for the case where
collisions are ignored.

energy lost when C R o 1. Its value is difcult to estimate from eld

work. For this reason, we study all possible values of CR from CR 0
(totally inelastic collision) to CR 1 (totally elastic collision).
According to the travel distances shown in Fig. 8 for various
restitution coefcients, one observes that particles travel farther
in the collision-enabled case than in collision-disabled case, even
when the restitution coefcient goes to zero. The travel distance is
observed to increase almost linearly with CR.
Relation between collisions and travel distance: As a general rule,
we have observed that a particle which travels far collides only a
few times before landing. Note that the total number of collisions
(M) and the average number of collisions per particle (m) have to
be distinguished. The relation between them is typically as
follows: with a small total number of collisions M, one has a small
number of particles that experience only one collision, thus m o 1.
As M increases, more particles will experience one collision
(m-1), up to the point where many particles experience many
collisions (m 4 2). That is why we argue that the deposition area
rst increases with M, and then decreases.

Fig. 5. Changes in the distribution of the deposition distance as a function of the

total number N of particles ejected by the volcano. Cases with and without collision
are shown.

Fig. 4. Illustration of the trajectories of a two-particle collision. The light particle (white circle) has a mass of 3 kg, and is ejected at time t 1 0 with speed 56 m/s. The heavy
one (black disk) has a mass of 81 kg and is ejected at time t 2 4:8 s, with speed 68 m/s. Panel (a) shows the trajectories before collision (solid lines) and the trajectories that
would be followed if collision is ignored (dashed lines). Panel (b) shows the positions of the particles at the moment of the collision, at time t 7.6 s. The dashed lines show
their trajectories after collision. Just before collision, the light particle has a velocity of 16.6 m/s and the heavy one a velocity of 41.1 m/s. After collision, the speed of the light
particle is 72.2 m/s and the speed of the heavy one is 38.7 m/s.

K. Tsunematsu et al. / Computers & Geosciences 63 (2014) 6269

Fig. 6. (a) Deposition distance versus burst time interval. (b) Number of different
group collisions versus burst time interval.

Our study also shows that a collision between a large and a

small particle is likely to increase the deposition distance of the
small particle. Accordingly, we have analyzed the relationship
between the number of collisions that a single particle experiences
and its deposition distance. We have also studied this relationship
as a function of the particle mass.
As shown in Fig. 9(a) the deposition distance decreases as a
given particle is subject to an increasing number of collisions. The
most favorable event to travel far is to experience one or two
collisions. Particles that collide more than 10 times do not go
farther than 200 m. Particles that deposit farther than 500 m
typically experienced less than 6 collisions.
The relationship between particle mass and deposition distance
for each particle is shown in Fig. 9(b). The deposition distance
decreases as the particle mass increases. The particle that traveled
the farthest has collided only once and has a mass 1.9 kg. Its travel
distance is 1900 m. In addition, we observe that the particles which
travel more than 500 m have a mass smaller than 50 kg. Particles
with a mass larger than 500 kg typically stay within 100 m of the
vent. This supports the interpretation that light particles travel
farther by receiving the large momentum from the heavy ones.
Without collisions, and provided that all particles are ejected
from the crater with the same speed, the mass distribution in the
landing area would be uniform. Therefore, if a non-uniform spatial
distribution of the mass of particles is observed around the crater,
as in Fig. 9(b), this could be a sign of the occurrence of airborne


Fig. 7. (a) Number of collisions versus particle diameter. (b) Changes in deposition
distance of collision-allowed cases and no-collision-allowed cases as a function of
the particle diameter.

Fig. 8. Changes in deposition distance for the cases with and without collisions as a
function of the restitution coefcient.

Note that some papers (e.g. Rosi et al., 2006; Wright et al.,
2007) proposed to estimate the velocities of ballistics from the
deposition distance. If collisions may occur during the ight in the
air, such a calculation can no longer be considered. Direct
measurement of the ejection speed is therefore recommended
unless one can exclude the presence of collisions.


K. Tsunematsu et al. / Computers & Geosciences 63 (2014) 6269

been observed that also Vulcanian/subplinian eruptions can

experience pulsating activity at vent. As an example, Ripepe
et al. (2013) have shown how the eruptive plume generated
during the Eyjafjallajkull 2010 eruption (Iceland) was characterized by discrete pulses injected into the atmosphere every 20 s for
56 s each. Pulsating eruptive behavior at source has already been
observed during other volcanic eruptions of various magma
compositions (e.g. Ruapehu, Bonadonna et al., 2005; Cerro Negro,
Hill et al., 1998; Etna, Vergniolle and Ripepe, 2008; Andronico
et al., 2008). However, it is expected that Strombolian particles
may experience a signicant drag force due to their vesicular
nature. Therefore, our model may well overestimate the actual
travel distance.

4. Conclusions
We have implemented three-dimensional ballistic simulations
taking into account the collisions between particles. As much as
possible, the parameters of the model are taken from eld
observations reported in the literature for Strombolian eruptions
at Stromboli volcano.
From this model, the particle distribution on the ground can be
computed and hazard maps, which accounts for inter-particles
collisions, can be produced.
To our knowledge, this is the rst study which investigates the
effect of collisions on the size of the deposition area around the crater
and performs a sensitivity analysis. As we assume parabolic trajectories for the bombs, we can compute analytically the collision
events and efciently implement our model with a DES method.
From our analysis of the simulation results, we conclude that

Fig. 9. (a) Deposition distance versus the number of collisions experienced by each
particle during a simulation. (b) Deposition distance versus the mass of each
particle. The results of ten simulation runs are merged in the plot.

3.3. Collision effect versus drag effect

Although we neglected the drag force in this study, some
comments can be made. By using the same simulation conditions
as before (see Table 2), we have computed that the travel distance
of a particle subject to a drag force is reduced by about 2050%,
depending on its size. On the other hand, the present study
(Figs. 58) demonstrated that collisions extend the travel distances
by a factor of two to six. From these gures, we can conclude that
collisions have a larger potential to modify the travel distance than
the drag force. When both collisions and drag affect the trajectories of the particles, we can argue that collisions will increase
the travel distance with respect to the situation where only the
drag force is present. However, this distance cannot be quantied
without performing detailed simulations. This will be considered
in a forthcoming paper.
3.4. Applicability of the model to various types of eruption
As discussed in Section 3.2, the most relevant parameter that
governs the collision rate is the time interval between bursts.
Consequently, Strombolian eruptions are the most likely explosive
events to be subject to inter-particle collisions. However, it is
expected that Strombolian fragments may experience a signicant
drag force due to their vesicular nature, in which case, our model
would overestimate the actual travel distance. Recently, it has

1. The collision probability is mostly affected by the burst time

interval. The most efcient collisions happen between particles
ejected in different bursts.
2. Effects of inter-particle collisions is most important for explosive volcanic eruptions characterized by pulsating behavior at
source (e.g. Strombolian, violent Strombolian, long-lasting
Vulcanian eruptions).
3. The travel distance of the particles increases with the total
number of collisions up to the point where individual particles
experience more than a few of them. Too many collisions per
particles have a neutralizing effects.
4. The mechanism that makes a particle travel farther when it
experiences a collision is the transfer of momentum that occurs
during the impact. The favorable events are those where the
speeds of the colliding particles form an angle close to 1801.
The effect is enhanced when the mass of the particle going up
is larger than the mass of the particle going down.
5. The spatial distribution of the mass of deposited particles is
affected by the presence of collisions. This distribution can then
be studied to assess the existence of collision during the
eruption. As a consequence, if collision is dominant, the
deposition distance cannot be used to derive important eruption parameters, such as the exit speed.
6. The increase of the travel distance is visible for any value of the
restitution coefcient. However, the increase is more pronounced when the kinetic energy is conserved (CR 1).
7. The effect of collisions on the travel distance of the particles is
as signicant in magnitude as the drag effect, but in an opposite
Alatorre-Ibargengoitia, M.A., Delgado-Granados, H., 2006. Experimental determination of drag coefcient for volcanic materials: calibration and application of a

K. Tsunematsu et al. / Computers & Geosciences 63 (2014) 6269

model to Popocatpetl volcano (Mexico) ballistic projectiles. Geophys. Res. Lett.

33 5 pp.
Alatorre-Ibargengoitia, M.A., Delgado-Granados, H., Dingwell, D.B., 2012. Hazard
map for volcanic ballistic impacts at Popocatpetl volcano (Mexico). Earth
Planet. Sci. Lett. 74, 21552169.
Andronico, D., Scollo, S., Caruso, S., Cristaldi, A., 2008. The 200203 Etna explosive
activity: Tephra dispersal and features of the deposits. J. Geophys. Res. 113.
Banks, J., Nelson, B., 2010. Discrete-Event System Simulation. Prentice Hall.
Blackburn, E., Wilson, L., Sparks, R., 1976. Mechanisms and dynamics of strombolian
activity. J. Geol. Soc. Lond. 132, 429440.
Bonadonna, C., Phillips, J., Houghton, B.F., 2005. Modeling Tephra sedimentation
from a Ruapehu weak plume eruption. J. Geophys. Res. 110, 122.
Bower, S.M., Woods, A.W., 1996. On the dispersal of clasts from volcanic craters
during small explosive eruptions. J. Volcanol. Geotherm. Res. 73, 1932.
Chouet, B., Hamisevicz, N., Mcgetchin, T.R., 1974. Photoballistics of volcanic jet
activity at Stromboll, Italy. J. Geophys. Res. 79, 49614976.
de' Michieli Vitturi, M., Neri, A., Ongaro, T.E., Savio, S.L., Boschi, E., 2010. Lagrangian
modeling of large volcanic particles: application to vulcanian explosions. J.
Geophys. Res. 115.
Fagents, S., Wilson, L., 1993. Explosive volcanic eruptions: VII. the ranges of
pyroclasts ejected in transient volcanic explosions. Geophys. J. Int. 113,
Hill, B.E., Connor, C.B., Jarzemba, M.S., La Femina, P.C., Navarro, M., Strauch, W.,
1998. 1995 Eruptions of Cerro Negro volcano, Nicaragua, and risk assessment
for future eruptions. Geol. Soc. Am. Bull. 110, 12311241.
Landau, L., Lifshitz, E., 1976. Mechanics 1. Course of Theoretical Physics. Elsevier


Mastin, L.G., 2002. A Simple Calculator of Ballistic Trajectories for Blocks Ejected
During Volcanic Eruptions. U.S. Geological Survey Open-File Report 01-45,
Version 1.1. U.S. Department of the Interior, U.S. Geological Survey.
Patrick, M.R., Harris, A.J.L., Ripepe, M., Dehn, J., Rothery, D.A., Calvari, S., 2007.
Strombolian explosive styles and source conditions: insights from thermal
(FLIR) video. Bull. Volcanol. 69, 769784.
Ripepe, M., Rossi, M., Saccorotti, G., 1993. Image processing of explosive activity at
Stromboli. J. Volcanol. Geotherm. Res. 54, 335351.
Ripepe, M., Bonadonna, C., Folch, A., Donne, D.D., Lacanna, G., Marchetti, E.,
Hskuldsson, A., 2013. Ash-plume dynamics and eruption source parameters
by infrasound and thermal imagery: the 2010 Eyjafjallajkull eruption. Earth
Planet. Sci. Lett. 366, 112121.
Rosi, M., Bertagnini, A., Harris, A.J.L., Pioli, L., Pistolesi, M., Ripepe, M., 2006. A case
history of paroxysmal explosion at Stromboli: timing and dynamics of the April
5, 2003 event. Earth Planet. Sci. Lett. 243, 594606.
Saunderson, H.C., 2008. Equations of motion and ballistic paths of volcanic ejecta.
Comput. Geosci. 34, 802814.
Vanderkluysen, L., Harris, A.J.L., Kelfoun, K., Bonadonna, C., Ripepe, M., 2012. Bombs
behaving badly: unexpected trajectories and cooling of volcanic projectiles.
Bullet. Volcanol. 74 (8), http://dx.doi.org/10.1007/s00445-012-0635-8.
Vergniolle, S., Ripepe, M., 2008. From strombolian explosions to re fountains at
Etna volcano (Italy): what do we learn from acoustic measurements?. Geol. Soc.
Lond. Spl. Publ. 307, 103124.
Wilson, L., 1972. Explosive volcanic eruptions-II the atmospheric trajectories of
pyroclasts. Geophys. J. R. Astron. Soc. 30, 381392.
Wright, H.M.N., Cashman, K.V., Rosi, M., Cioni, R., 2007. Breadcrust bombs as
indicators of vulcanian eruption dynamics at Guagua Pichincha volcano,
Ecuador. Bull. Volcanol. 69, 281300.